NewQUORUM: a team of AI agents on every pentest

The AI pentester that never guesses.

AIpento sends a team of AI agents at your running app: 155 live attacks, every finding reproduced with the exact request and response, every fix verified twice. Scanners guess. AIpento proves.

Free readiness scans, foreverAI pentest on ProOnly against targets you authorize

  1. 01DiscoverAI agents map every endpoint.
  2. 02AttackLive probes fire at each one.
  3. 03BreachOne lands. The evidence is kept.
  4. 04FixA patch the gate re-scans.
  5. 05VerifiedGreen twice. Only then fixed.
326static checks on every scan
155live probes against your running app
14attack vectors, OWASP Web, API and LLM
255,163known-vulnerability advisories screened (OSV)
2×your suite passes before a fix is accepted

Live attack console

Watch it break your app. Then read the evidence.

Every probe you see firing is a real check from the engine’s catalogue, in the order it runs them. When one lands you get the exact request, the response that came back, and why it matters.

aipento attack --target vuln-fixture.local --authorizedrunning

Probes 0 / 155

A replay of a real engine run against our own deliberately vulnerable test app, on localhost, built to trip every probe. Every probe and every finding shown is the engine’s actual output. For the report you receive, see the sample report.

Evidence, not a score

Reproduced, never assumed.

  1. 01

    A finding only counts when it runs

    We send the request, capture the response and keep the transcript. Nothing reaches your report on the strength of a pattern match alone.

  2. 02

    Safe by construction

    Pentests run only against a target you authorize in writing. Never production, never a system you did not name. Scans only read your code.

  3. 03

    Accepted on green, twice

    A fix is not trusted because a model wrote it. The gate re-scans it clean, then your own suite passes twice from a clean checkout.

CCriticalS2 · security lens

SQL built by joining strings

api/tasks.js:7

sql: "SELECT * FROM tasks WHERE title LIKE '%" + q + "%'",
Why it matters
A value from a visitor becomes part of the query. This is how a form field turns into “read every row in the table”.
What to do
Use parameterised queries. Never concatenate or interpolate a value into SQL.
From the Taskly sample report · open it

Attack surface

See the whole surface, and how the holes connect.

AIpento maps what your app exposes, what was attacked and where it broke, left to right from the app to the finding. Where findings could chain into a real attack path, it draws the chain and labels it inferred until a probe proves it.

Illustration of the Attack Surface view. Vector names are the engine’s real ones.

AutoFix

Fixes that have to prove themselves.

Send any finding to AutoFix. Your configured models are asked in turn for a patch. One is accepted only when the gate re-scans it clean, and it is marked verified twice once your suite passes twice from a clean checkout.

  1. Draftyour models, in turn
  2. Adjudicateone patch wins
  3. Verifythe gate re-scans clean
  4. Lockgreen twice
api/tasks.js+2-1 Verified twice
  export async function search(q) {
-   sql: "SELECT * FROM tasks WHERE title LIKE '%" + q + "%'",
+   sql: "SELECT * FROM tasks WHERE title LIKE ?",
+   args: [`%${q}%`],
  }
Illustrative patch for the Taskly S2 finding.

Why this exists

Building got cheap. Knowing it is safe did not.

45%

of code-generation tasks introduced a known security flaw, across more than 150 models.

Veracode, GenAI Code Security, Spring 2026
85%

of the time, the models failed to defend against cross-site scripting.

Veracode, same study
10×

more security findings a month, about 1,000 to over 10,000 in six months, as AI-assisted teams committed faster.

Apiiro, via the Cloud Security Alliance, April 2026

The full catalogue

Every check has a name. Every one can fail you.

326 static checks across 13 lenses and 155 live probes across 14 attack vectors, plus every dependency screened against 255,163 OSV advisories. Search it.

Live probes

Transport & headers3
  • Missing security response headers
  • Cross-origin resource sharing misconfigured
  • Session cookie missing security flags
Exposure2
  • Secret exposed in a response
  • Verbose error or stack trace disclosure
Abuse & limits1
  • No rate limiting on authentication
SSRF & out-of-band2
  • Server-side request forgery
  • Blind server-side request forgery
Recon5
  • Sensitive file served to the public
  • JavaScript source map served in production
  • API documentation publicly reachable
  • GraphQL introspection enabled
  • Dangerous HTTP method accepted unauthenticated
Advanced4
  • Open redirect
  • Path traversal
  • Account enumeration
  • Forwarded-host header trusted for a password-reset link
Exposed files28
  • AWS credentials file publicly reachable
  • AWS CLI config publicly reachable
  • SSH private key publicly reachable
  • SSH private key in .ssh publicly reachable
  • npm registry auth token publicly reachable
  • Apache htpasswd hash file publicly reachable
  • Production environment file publicly reachable
  • Local environment file publicly reachable
  • .NET appsettings with connection strings publicly reachable
  • IIS web.config publicly reachable
  • WordPress config backup publicly reachable
  • Apache Airflow configuration publicly reachable
  • Ansible configuration publicly reachable
  • AWS Elastic Beanstalk Docker config publicly reachable
  • Docker Compose file publicly reachable
  • GitLab CI pipeline publicly reachable
  • CircleCI config publicly reachable
  • Travis CI config publicly reachable
  • phpinfo() output publicly reachable
  • Apache server-status page publicly reachable
  • Spring Boot actuator env publicly reachable
  • Spring Boot actuator health publicly reachable
  • Prometheus metrics endpoint publicly reachable
  • Go expvar debug endpoint publicly reachable
  • Laravel application log publicly reachable
  • VS Code SFTP credentials publicly reachable
  • Firebase project config publicly reachable
  • Rails database.yml with credentials publicly reachable
Access control18
  • Session id not rotated on login (session fixation)
  • Session not invalidated server-side on logout
  • Session cookie has an over-long or absent lifetime
  • JWT algorithm confusion (RS256 verified as HS256 with the public key)
  • JWT kid header injection selects an attacker-controlled key
  • Unsigned JWT (alg:none) accepted
  • CORS reflects a null origin with credentials
  • Cross-tenant read via a tenant header
  • Forced browsing to an admin-only sub-function as a normal user
  • Horizontal escalation by swapping an owner id in the POST body
  • Password-reset token is guessable and/or replayable
  • Sensitive action performed without re-authentication
  • Broken object-level access control (IDOR)
  • Protected endpoint reachable without authentication
  • Authenticated cross-account access (BOLA)
  • Function-level privilege escalation (BFLA)
  • Mass assignment on profile update
  • alg:none JWT accepted
Injection15
  • Error-based SQL injection
  • Boolean-based blind SQL injection
  • Time-based blind SQL injection
  • NoSQL operator injection
  • OS command injection
  • Server-side template injection
  • XML external entity expansion
  • CRLF / response-header injection
  • GraphQL query batching / amplification
  • GraphQL field-suggestion leakage
  • Mass assignment on a nested object
  • Stored cross-site scripting
  • XSS via a value echoed into a script sink
  • Reflected cross-site scripting
  • SQL injection
AI / LLM surface10
  • Direct prompt injection overrides the system instruction
  • Indirect prompt injection via an ingested document
  • System prompt / instruction leak
  • Insecure output handling: model output rendered as raw HTML
  • Agent tool call runs outside the caller scope (confused deputy)
  • Unbounded token / cost per request
  • Secret / PII disclosed in a completion
  • RAG retrieval returns another tenant's data
  • Refusal bypass reaches a privileged tool
  • AI endpoint unauthenticated or prompt-injectable
Exposed admin panels18
  • Exposed Grafana panel
  • Exposed Kibana panel
  • Exposed Jenkins panel
  • Exposed phpMyAdmin panel
  • Exposed Adminer panel
  • Exposed Prometheus panel
  • Exposed Apache Airflow panel
  • Exposed Django admin panel
  • Exposed Spring Boot Actuator panel
  • Exposed Werkzeug debug console panel
  • Exposed admin login with a default-credential hint panel
  • Exposed Kubernetes Dashboard panel
  • Exposed RabbitMQ Management panel
  • Exposed Swagger/OpenAPI UI panel
  • Exposed Traefik dashboard panel
  • Exposed Redis Commander panel
  • Exposed pgAdmin panel
  • Exposed MinIO Console panel
Misconfiguration18
  • Directory listing enabled
  • Verbose debug endpoint dumps a stack trace
  • Permissive crossdomain.xml
  • Permissive clientaccesspolicy.xml
  • Missing Strict-Transport-Security
  • CSP weakened by 'unsafe-inline'
  • Dangerous HTTP method advertised via Allow
  • Cookie SameSite=None without Secure
  • Page is framable (no anti-clickjacking control)
  • CORS preflight reflects any origin with credentials
  • Missing X-Content-Type-Options: nosniff
  • Server banner leaks exact software versions
  • Sensitive JSON response is cacheable
  • Internal config exposed under /.well-known/
  • Referrer-Policy leaks full URLs
  • Permissive Access-Control-Expose-Headers
  • Source map exposed from a production bundle
  • Mixed content (http resources on a secure page)
LLM abuse & poisoning17
  • User-submitted "correction" is trusted as authoritative ground truth
  • One request triggers unbounded expansion with no server-side cap
  • Endpoint runs an unpinned model from an untrusted source
  • Tool accepts free-form arguments that reach a sensitive action
  • Agent chains tools beyond the caller's granted scope
  • Agent can self-invoke with no recursion cap
  • Factual compliance claim returned with no citation or guardrail
  • Systematic model extraction is not rate limited and exposes internals
  • Injection accumulated across turns bypasses a single-turn filter
  • Base64-encoded instruction slips a plaintext guardrail
  • Indirect injection via the content of a retrieved URL
  • Tool output is trusted verbatim as instructions
  • Conversation memory leaks across sessions
  • Agent tool fetches an attacker-supplied URL (SSRF)
  • Model-written code executes unsandboxed and reaches the host
  • Compliance decision fails open to "approved" when the model errors
  • Prompts are logged verbatim and retrievable unscrubbed
API abuse14
  • BOLA on a second object type (invoices) read by id
  • Bulk endpoint returns objects for ids the caller does not own
  • Endpoint returns more fields than the UI needs, including PII
  • No page-size cap on an expensive list endpoint
  • A normal user reaches an admin-only API function
  • No anti-automation on a sensitive business flow (coupon redemption)
  • Verbose API error leaks a stack trace and the SQL
  • A debug header switches on internal diagnostics disclosure
  • A legacy /apix/v1 endpoint is still live and weaker than v2
  • Webhook register fetches an attacker-supplied URL (SSRF to internal)
  • Deeply nested query accepted with no complexity limit
  • An access token is placed in a shareable URL query string
  • Walking pagination offsets returns other tenants' records
  • Mass export returns records beyond the caller's tenant

Static checks

Tests that can failT1-T4
  • No automated tests anywhere in the codebase
  • Test file contains no assertions
  • Focused or skipped test committed
  • Tests exist but nothing runs them automatically
Security reviewS1-S22
  • Live credential committed to the codebase
  • SQL built by joining strings
  • Untrusted value written into the page as HTML
  • Code built and executed at runtime
  • Endpoint that writes data with no visible authentication
  • Cross-origin requests allowed from anywhere with credentials
  • Administrative database key referenced in client-side code
  • Password hashed with a fast, unsalted algorithm
  • Bcrypt cost factor set below a safe minimum
  • Math.random used to generate a token, secret or session id
  • JWT verification accepts an unsigned or already-expired token
  • Redirect target taken directly from request input
  • Outbound request built directly from request input
  • File path built directly from request input
  • Shell command built directly from request input
  • TLS certificate verification disabled
  • Request body written directly into a create or update call
  • Secret-looking name exposed through a browser-bundled environment variable
  • Stripe webhook handler with no signature verification
  • Error stack trace returned in the HTTP response body
  • Untrusted data deserialized with pickle or an unsafe YAML loader
  • Login credentials or session token written to the log
Data you can rebuildD1-D4
  • Database in use with no migration history
  • Environment file is not excluded from version control
  • Table created in a Supabase migration with no row-level security
  • Security rules allow unrestricted read and write
Failure pathsF1-F3
  • Error caught and discarded
  • Network call with no error handling in the file
  • Network call with no timeout
Speed at real sizeP1-P3
  • Query with no limit on how many rows come back
  • Database query inside a loop
  • Synchronous file call blocks a request handler
Release and rollbackR1-R4
  • No dependency lockfile
  • Dependency pinned to a moving target
  • No deployment configuration in the repository
  • Production build configured to ship source maps
Agent tools and MCP serversM1-M7
  • Agent tool runs shell commands built from model input
  • File tool builds a path from model input with no root confinement
  • Tool fetches a URL chosen by the model with no allowlist
  • Tools registered with no input schema
  • Content fetched from elsewhere is returned straight into the model context
  • Destructive or spending tool with no confirmation step
  • Tool arguments written to logs
Committed secretsSEC-1PASSWORD-SECRET-KEY-SEC-ZENDESK-SECRET-KEY
  • Committed credential: 1password Secret Key
  • Committed credential: 1password Service Account Token
  • Committed credential: Adafruit Api Key
  • Committed credential: Adobe Client Id
  • Committed credential: Adobe Client Secret
  • Committed credential: Age Secret Key
  • Committed credential: Airtable Api Key
  • Committed credential: Airtable Personnal Access Token
  • Committed credential: Algolia Api Key
  • Committed credential: Alibaba Access Key Id
  • Committed credential: Alibaba Secret Key
  • Committed credential: Anthropic Admin Api Key
  • Committed credential: Anthropic Api Key
  • Committed credential: Artifactory Api Key
  • Committed credential: Artifactory Reference Token
  • Committed credential: Asana Client Id
  • Committed credential: Asana Client Secret
  • Committed credential: Aws Access Token
  • Committed credential: Aws Amazon Bedrock Api Key Long Lived
  • Committed credential: Aws Amazon Bedrock Api Key Short Lived
  • Committed credential: Azure Ad Client Secret
  • Committed credential: Beamer Api Token
  • Committed credential: Bitbucket Client Id
  • Committed credential: Bitbucket Client Secret
  • Committed credential: Bittrex Access Key
  • Committed credential: Bittrex Secret Key
  • Committed credential: Clickhouse Cloud Api Secret Key
  • Committed credential: Clojars Api Token
  • Committed credential: Cloudflare Api Key
  • Committed credential: Cloudflare Global Api Key
  • Committed credential: Cloudflare Origin Ca Key
  • Committed credential: Codecov Access Token
  • Committed credential: Coinbase Access Token
  • Committed credential: Confluent Access Token
  • Committed credential: Confluent Secret Key
  • Committed credential: Contentful Delivery Api Token
  • Committed credential: Curl Auth Header
  • Committed credential: Curl Auth User
  • Committed credential: Databricks Api Token
  • Committed credential: Datadog Access Token
  • Committed credential: Defined Networking Api Token
  • Committed credential: Digitalocean Access Token
  • Committed credential: Digitalocean Pat
  • Committed credential: Digitalocean Refresh Token
  • Committed credential: Discord Api Token
  • Committed credential: Discord Client Id
  • Committed credential: Discord Client Secret
  • Committed credential: Doppler Api Token
  • Committed credential: Droneci Access Token
  • Committed credential: Dropbox Api Token
  • Committed credential: Dropbox Long Lived Api Token
  • Committed credential: Dropbox Short Lived Api Token
  • Committed credential: Duffel Api Token
  • Committed credential: Dynatrace Api Token
  • Committed credential: Easypost Api Token
  • Committed credential: Easypost Test Api Token
  • Committed credential: Facebook Access Token
  • Committed credential: Facebook Page Access Token
  • Committed credential: Facebook Secret
  • Committed credential: Fastly Api Token
  • Committed credential: Finicity Api Token
  • Committed credential: Finicity Client Secret
  • Committed credential: Finnhub Access Token
  • Committed credential: Flickr Access Token
  • Committed credential: Flutterwave Encryption Key
  • Committed credential: Flutterwave Public Key
  • Committed credential: Flutterwave Secret Key
  • Committed credential: Flyio Access Token
  • Committed credential: Frameio Api Token
  • Committed credential: Freemius Secret Key
  • Committed credential: Freshbooks Access Token
  • Committed credential: Gcp Api Key
  • Committed credential: Github App Token
  • Committed credential: Github Fine Grained Pat
  • Committed credential: Github Oauth
  • Committed credential: Github Pat
  • Committed credential: Github Refresh Token
  • Committed credential: Gitlab Cicd Job Token
  • Committed credential: Gitlab Deploy Token
  • Committed credential: Gitlab Feature Flag Client Token
  • Committed credential: Gitlab Feed Token
  • Committed credential: Gitlab Incoming Mail Token
  • Committed credential: Gitlab Kubernetes Agent Token
  • Committed credential: Gitlab Oauth App Secret
  • Committed credential: Gitlab Pat
  • Committed credential: Gitlab Pat Routable
  • Committed credential: Gitlab Ptt
  • Committed credential: Gitlab Rrt
  • Committed credential: Gitlab Runner Authentication Token
  • Committed credential: Gitlab Runner Authentication Token Routable
  • Committed credential: Gitlab Scim Token
  • Committed credential: Gitlab Session Cookie
  • Committed credential: Gitter Access Token
  • Committed credential: Gocardless Api Token
  • Committed credential: Grafana Api Key
  • Committed credential: Grafana Cloud Api Token
  • Committed credential: Grafana Service Account Token
  • Committed credential: Harness Api Key
  • Committed credential: Hashicorp Tf Password
  • Committed credential: Heroku Api Key
  • Committed credential: Heroku Api Key V2
  • Committed credential: Hubspot Api Key
  • Committed credential: Infracost Api Token
  • Committed credential: Intercom Api Key
  • Committed credential: Intra42 Client Secret
  • Committed credential: Jfrog Api Key
  • Committed credential: Jfrog Identity Token
  • Committed credential: Jwt
  • Committed credential: Kraken Access Token
  • Committed credential: Kucoin Access Token
  • Committed credential: Kucoin Secret Key
  • Committed credential: Launchdarkly Access Token
  • Committed credential: Linear Api Key
  • Committed credential: Linear Client Secret
  • Committed credential: Linkedin Client Id
  • Committed credential: Linkedin Client Secret
  • Committed credential: Lob Api Key
  • Committed credential: Lob Pub Api Key
  • Committed credential: Looker Client Id
  • Committed credential: Looker Client Secret
  • Committed credential: Mailchimp Api Key
  • Committed credential: Mailgun Private Api Token
  • Committed credential: Mailgun Pub Key
  • Committed credential: Mailgun Signing Key
  • Committed credential: Mapbox Api Token
  • Committed credential: Mattermost Access Token
  • Committed credential: Maxmind License Key
  • Committed credential: Messagebird Api Token
  • Committed credential: Messagebird Client Id
  • Committed credential: Microsoft Teams Webhook
  • Committed credential: Netlify Access Token
  • Committed credential: New Relic Browser Api Token
  • Committed credential: New Relic Insert Key
  • Committed credential: New Relic User Api Id
  • Committed credential: New Relic User Api Key
  • Committed credential: Notion Api Token
  • Committed credential: Npm Access Token
  • Committed credential: Nuget Config Password
  • Committed credential: Nytimes Access Token
  • Committed credential: Octopus Deploy Api Key
  • Committed credential: Openai Api Key
  • Committed credential: Openshift User Token
  • Committed credential: Perplexity Api Key
  • Committed credential: Plaid Api Token
  • Committed credential: Plaid Client Id
  • Committed credential: Plaid Secret Key
  • Committed credential: Planetscale Api Token
  • Committed credential: Planetscale Oauth Token
  • Committed credential: Planetscale Password
  • Committed credential: Postman Api Token
  • Committed credential: Prefect Api Token
  • Committed credential: Private Key
  • Committed credential: Pulumi Api Token
  • Committed credential: Pypi Upload Token
  • Committed credential: Rapidapi Access Token
  • Committed credential: Readme Api Token
  • Committed credential: Rubygems Api Token
  • Committed credential: Scalingo Api Token
  • Committed credential: Sendbird Access Id
  • Committed credential: Sendbird Access Token
  • Committed credential: Sendgrid Api Token
  • Committed credential: Sendinblue Api Token
  • Committed credential: Sentry Access Token
  • Committed credential: Sentry Org Token
  • Committed credential: Sentry User Token
  • Committed credential: Settlemint Application Access Token
  • Committed credential: Settlemint Personal Access Token
  • Committed credential: Settlemint Service Access Token
  • Committed credential: Shippo Api Token
  • Committed credential: Shopify Access Token
  • Committed credential: Shopify Custom Access Token
  • Committed credential: Shopify Private App Access Token
  • Committed credential: Shopify Shared Secret
  • Committed credential: Sidekiq Secret
  • Committed credential: Sidekiq Sensitive Url
  • Committed credential: Slack App Token
  • Committed credential: Slack Bot Token
  • Committed credential: Slack Config Access Token
  • Committed credential: Slack Config Refresh Token
  • Committed credential: Slack Legacy Bot Token
  • Committed credential: Slack Legacy Token
  • Committed credential: Slack Legacy Workspace Token
  • Committed credential: Slack User Token
  • Committed credential: Slack Webhook Url
  • Committed credential: Snyk Api Token
  • Committed credential: Sonar Api Token
  • Committed credential: Sourcegraph Access Token
  • Committed credential: Square Access Token
  • Committed credential: Squarespace Access Token
  • Committed credential: Stripe Access Token
  • Committed credential: Travisci Access Token
  • Committed credential: Twilio Api Key
  • Committed credential: Twitch Api Token
  • Committed credential: Twitter Access Secret
  • Committed credential: Twitter Access Token
  • Committed credential: Twitter Api Key
  • Committed credential: Twitter Api Secret
  • Committed credential: Twitter Bearer Token
  • Committed credential: Typeform Api Token
  • Committed credential: Vault Batch Token
  • Committed credential: Yandex Access Token
  • Committed credential: Yandex Api Key
  • Committed credential: Yandex Aws Access Token
  • Committed credential: Zendesk Secret Key
Known-vulnerable dependenciesDEP-CVE-DEP-CVE
  • Dependency pinned to a version with a known published vulnerability
Container, CI & infrastructureIF1-IF19
  • Container image runs as the root user
  • Base image pinned to a mutable or latest tag
  • Secret baked into a Dockerfile ENV or ARG
  • Dockerfile pipes a downloaded script straight into a shell
  • Dockerfile uses ADD for a remote URL or archive
  • Compose service granted full privileged mode
  • Compose service binds the host network or every interface
  • Secret hardcoded in a Compose environment value
  • GitHub Action pinned to a mutable reference
  • Workflow checks out PR code under pull_request_target
  • Workflow step prints a secret to the log
  • Workflow interpolates untrusted event text into a run script
  • Security group opens a sensitive port to the whole internet
  • S3 bucket configured for public access
  • Storage or database resource created with encryption disabled
  • Secret hardcoded in a Terraform configuration
  • Kubernetes container runs privileged or can escalate privileges
  • Kubernetes pod uses host networking or a host path volume
  • Kubernetes container image uses latest or no tag
Python securityPY1-PY22
  • YAML parsed with an unsafe loader
  • Subprocess launched through a shell
  • Flask development server started with the debugger enabled
  • Assert statement used as a runtime authorization check
  • Weak hash algorithm used for a digest
  • TLS certificate verification disabled on an HTTP client
  • Django settings ship with DEBUG enabled
  • Hardcoded Django SECRET_KEY committed to the codebase
  • Code built and executed at runtime with eval or exec
  • Insecure temporary file created with tempfile.mktemp
  • XML parsed without protection against entity expansion
  • Jinja2 environment created with autoescaping turned off
  • Operating-system command run through os.system
  • Service bound to all network interfaces
  • Untrusted value marked safe for HTML rendering
  • Security token generated with the non-cryptographic random module
  • SQL query assembled with string formatting before execute
  • SSH host-key checking disabled with AutoAddPolicy
  • Weak hash algorithm selected through hashlib.new
  • TLS verification disabled through an unverified SSL context
  • Archive extracted without validating member paths
  • Django ALLOWED_HOSTS opened to every host
Go securityGO1-GO17
  • SQL query built with fmt.Sprintf or string concatenation
  • Shell command run through sh -c or a variable command
  • TLS certificate verification disabled with InsecureSkipVerify
  • Weak hash md5 or sha1 used in application code
  • Token or secret generated with math/rand
  • SSH host key checking disabled with InsecureIgnoreHostKey
  • File or directory created with world-writable permissions
  • File path built from request input without containment
  • HTTP server started without read or write timeouts
  • Untrusted value wrapped in template.HTML
  • Redirect target taken from request input
  • Weak cipher DES or RC4 used for encryption
  • Request body read with no size limit
  • Executable run from a temporary or downloaded path
  • TLS minimum version pinned below 1.2
  • Service bound to all network interfaces
  • RSA key generated with fewer than 2048 bits
PHP securityPHP1-PHP16
  • SQL query built from request input or an interpolated variable
  • Request value echoed into the page without escaping
  • Code executed at runtime with eval on a variable
  • Shell command built from a variable
  • File included from request input
  • Untrusted input passed to unserialize
  • Password hashed with md5 or sha1
  • Removed ext/mysql function still in use
  • File or URL opened from request input
  • Uploaded file moved with no type allowlist
  • Secret compared with == instead of hash_equals
  • PHP configured to display errors to the browser
  • Request superglobal expanded with extract
  • HTTP header built from request input
  • Regex replacement using the code-executing e modifier
  • Curl TLS certificate verification disabled

Pricing

Free to see it. Pro to break it and fix it.

Free

$0

Connect a repo and run readiness scans yourself, forever.

  • Unlimited readiness scans
  • 326 checks, ranked Critical to Low with evidence
  • Live scan console and posture Command Center
  • A plain verdict: ship, fix first, or rebuild
Start free

Most teams

Pro

From$2,999/ month

Add the AI pentest and AutoFix, on your own projects, on demand.

  • Everything in Free
  • AI pentest: 155 live probes, live attack console
  • AutoFix: multi-model, gate-verified, retested
  • API keys and CI integration
Start free, upgrade in-app

Enterprise

Custom

Many apps, SSO, private runners, and a human-led fix sprint when you want one.

  • Unlimited projects and seats
  • SSO (OIDC), TOTP two-factor, four org roles
  • Tamper-evident audit log and auditor Evidence Pack
  • Self-hosted scan runners, on-prem edition
  • Fixed-price fix sprint: half now, half on green
Talk to us

Questions

What security teams ask first.

What is an AI pentest?

AI agents map your app, plan attacks the way a tester would, and run them against a staging URL you own and have authorized in writing. Every finding comes with the request and response that proved it. A finding the agents cannot reproduce is withdrawn, not reported.

What access do you need?

Read-only access to a repository for scans. For a pentest, a staging URL you own and a written authorization naming it. We never need production credentials, and nothing runs against a system you did not name.

How is this different from a scanner?

A scanner reports what a pattern looks like. AIpento runs the attack and keeps the request and response. If a finding does not reproduce, it does not reach your report.

What does “verified twice” mean?

A fresh checkout, a fresh database, the full suite, in your CI. Then the same again. Suites that pass once often lean on leftover data or test order. The second clean run is what separates a fix from a lucky run.

Can it run inside our network?

Yes. Enterprise includes self-hosted runners and an on-prem edition, with results in the same console as your cloud projects.

Which stacks do you cover?

Best fit today: React and Next.js, Node, Python, and Postgres or Supabase. If yours is something else, ask. We will say plainly if it is outside what we can verify well.

Enterprise

Private runners, SSO, or a managed fix sprint.

Tell us about the app. We reply with next steps, and a fixed price before any work starts. No call required.

  • Scope and price written down before kickoff
  • Half starts the sprint, half only on green
  • If we do not get you there, you owe nothing further

Start here

Find out what breaks before someone else does.

Readiness scans are free, forever. Add the AI pentest when you are ready to be attacked for real.